欢迎访问作物学报,今天是

作物学报 ›› 2021, Vol. 47 ›› Issue (2): 294-304.doi: 10.3724/SP.J.1006.2021.04085

• 作物遗传育种·种质资源·分子遗传学 • 上一篇    下一篇

运用广义线性混合模型分析随机区组重复测量的试验资料

张久权1,*(), 闫慧峰1, 褚继登1, 李彩斌2   

  1. 1中国农业科学院烟草研究所 / 农业农村部烟草生物学与加工重点实验室, 山东青岛 266101
    2贵州省烟草公司毕节市公司, 贵州毕节 551700
  • 收稿日期:2020-04-01 接受日期:2020-07-02 出版日期:2021-02-12 网络出版日期:2020-07-15
  • 通讯作者: 张久权
  • 基金资助:
    国家重点研发计划项目(2018YFD201104);四川省烟草公司重点项目(SCYC201702);四川省烟草公司凉山州公司项目(LSYC201601);贵州省烟草公司毕节市公司项目(2018520500240059)

Statistical analysis of randomized complete block design with repeated measure data using Generalized Linear Mixed Models (GLIMMIX)

ZHANG Jiu-Quan1,*(), YAN Hui-Feng1, CHU Ji-Deng1, LI Cai-Bin2   

  1. 1Tobacco Research Institute, Chinese Academy of Agriculture Sciences / Key Laboratory of Tobacco Biology and Processing, Ministry of Agriculture and Rural Affairs, Qingdao 266101, Shandong, China
    2Bijie Tobacco Company of Guizhou Province, Bijie 551700, Guizhou, China
  • Received:2020-04-01 Accepted:2020-07-02 Published:2021-02-12 Published online:2020-07-15
  • Contact: ZHANG Jiu-Quan
  • Supported by:
    National Key Research and Development Program of China(2018YFD201104);Sichuan Provincial Tobacco Company(SCYC201702);Liangshan Tobacco Company of Sichuan Province(LSYC201601);Bijie Tobacco Company of Guizhou Province(2018520500240059)

摘要:

重复测量试验对同一受试对象进行多次测量, 各时间点数据间存在自相关性, 进行方差分析和均值比较时需要进行特殊处理。虽然此方法在农业等研究领域运用十分广泛, 但目前有效地相关统计方法鲜见。为了建立操作简单、实用性强、结果可靠的统计分析方法, 本研究采用SAS的广义线性混合模型(Generalized Linear Mixed Models, GLIMMIX), 以随机区组重复测量试验资料为例, 说明了协方差结构筛选、方差分析和均值比较的具体方法。结果表明, 用传统的裂区设计、多变量统计等方法会造成资料信息浪费, 统计功效降低, 缺区无法处理等问题, 甚至会导致错误的结论。GLIMMIX能很好地处理自相关问题, 功能强大, 结果可靠, 使用简单, 允许缺区, 是进行重复测量试验资料方差分析和均值比较的理想方法。目前在国内将其运用到农学类试验数据的统计分析的相关报道鲜见, 该文在本领域具有很强的实用性和创新性。

关键词: 重复测量, 随机区组, 广义线性混合模型, 方差分析, 均值比较, SAS

Abstract:

Multiple measurements of the same subject are conducted, and there is autocorrelations among the data at each time point. Some special treatment is required for statistical analysis of repeated measure data. Although the repeated measure is widely used in agricultural and other research fields, the relevant and effective statistical methods are rare. In order to establish a simple, easy to use, and reliable statistical method, generalized linear mixed models (GLIMMIX) of SAS was adapted. Selection of covariance structure, variance analysis, and means comparison processes were showed by using RCB data. Traditional split plot and MANOVA methods wasted large amounts of information, reduced the power of the test, and could not handle missing data effectively, even resulting in incorrect conclusions. GLIMMIX was the best choice for variance analysis and means comparison of repeated measure data, because it was easy to use, and had powerful function, high reliability, and ability to handle missing data. At present, there was few relevant report in China, and this method would be very practical and innovative in this field.

Key words: repeated measure, randomized complete block, GLIMMIX, analysis of variance, mean comparison, SAS

表1

数据Microsoft Excel输入表示例"

Rain N Block Core Times Y
1 1 1 1 6 20.20
1 1 1 1 7 7.37
1 1 1 1 8 5.75
1 1 1 1 9 4.25
1 1 1 1 10 2.42
1 1 1 1 11 4.69
1 1 1 1 12 5.13
1 1 2 2 6 20.45
1 1 2 2 7 12.33
3 3 3 27 12 6.16

图1

3种施肥处理总氮淋失量随淋洗次数的变化 误差线为标准误。"

表4

F检验结果(III型, ANTE1, 土柱试验)"

效应
Effect
分子自由度
Numerator DF
分母自由度
Denominator DF
F
F-value
P
P-value
Rain 2 28.62 30.49 <0.0001
N 2 28.62 22.89 <0.0001
Rain*N 4 28.62 1.34 0.2806
Times 6 29.40 36.54 <0.0001
Rain*Times 12 39.27 0.90 0.5509
N*Times 12 39.27 4.21 0.0003
Rain*N*Times 24 47.73 0.48 0.9725

表5

处理均值两两比较示例"

语句#
Code #
处理组合
Treatment
差值
Difference
标准误
SE
自由度
DF
t
t-value
Pr > |t|
(4a) (1)定位法Positional 8.05 4.32 17.85 1.87 0.0786
(4b) (1)非定位法Non-positional 8.05 4.32 17.85 1.87 0.0786
(5) (2) 10.94 3.19 20.36 3.43 0.0026
(6) (3) 6.38 1.73 33.37 3.70 0.0008
(7) (4) 0.94 0.60 31.51 1.58 0.1248
(8) (5) 1.78 0.77 28.62 2.31 0.0285

表2

采用不同协方差结构模型时F检验P值(III型检验)"

效应
Effect
方差分量
VC
复合对称
CS
不规则
UN
空间幂相关
SP
一阶自回归AR(1) 循环相关
TOEP
一阶前依赖ANTE(1)
Rain <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
N <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.0001 <0.0001
Rain*N 0.2375 0.2609 0.2809 0.3023 0.3023 0.3161 0.2822
Times <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001
Rain*Times 0.0033 0.0038 0.6167 0.0059 0.0059 0.0172 0.5533
N*Times <0.0001 <0.0001 0.0017 <0.0001 <0.0001 <0.0001 0.0003
Rain*N*Times 0.8526 0.8537 0.9457 0.8584 0.8584 0.9225 0.9727

表3

不同协方差结构模型拟合性(土柱试验)"

准则
Criteria
方差分量
VC
复合对称
CS
不规则
UN
空间幂相关
SP
一阶自回归AR(1) 循环相关
TOEP
一阶前依赖ANTE(1)
-2logL 790.7 790.7 638.5 790.4 790.4 677.0 790.7
AIC 792.7 794.7 694.5 794.4 794.4 703.0 792.7
AICC 792.7 794.8 711.3 794.5 794.5 706.3 792.7
BIC 794.0 797.2 730.8 797.0 792.6 691.3 794.0
CAIC 795.0 799.2 758.8 799.0 794.6 704.3 795.0
HQIC 793.0 795.4 705.3 795.2 790.8 679.5 793.0

图2

方差分量结构"

[1] Wang Z, Goonewardene L A. The use of MIXED models in the analysis of animal experiments with repeated measures data. Can J Animal Sci, 2004,84:1-11.
[2] Kaps M, Lamberson W. Biostatistics for Animal Science, 3rd edn. Wallingford: CABI Publishing, 2017. pp 365-380.
[3] Littell R C, Henry P R, Ammerman C B. Statistical analysis of repeated measures data using sas procedures. J Animal Sci, 1998,76:1216-1231.
[4] SAS Institute. SAS Technique Support. Cary, NC, USA: SAS Institute Inc., 2020. [2020-05-02]. http://support.sas.com/techsup/.
[5] Littell R C, Milliken G A, Stroup W W, Wolfinger R D, Schabenberger O. SAS for Mixed Models, 2nd edn. Cary, NC, USA: SAS Institute Inc., 2006. pp 159-202.
[6] Yossa R, Verdegem M. Misuse of multiple comparison tests and underuse of contrast procedures in aquaculture publications. Aquaculture, 2015,437:344-350.
[7] 王琪, 胡良平, 高辉. 如何用SAS软件正确分析生物医学科研资料: XV. 用SAS软件实现具有一个重复测量的两因素和具有两个重复测量的两因素设计定量资料的统计分析. 中国医药生物技术, 2012,7(1):74-77.
Wang Q, Hu L P, Gao H. How to use SAS software to correctly analyze data of biomedical study: XV. Statistical analysis of quantitative data with one or two factors of repeated measure. Chin Med Biotechnol, 2012,7(1):74-77 (in Chinese).
[8] 周倩, 张晋昕. 含缺失值的重复测量资料分析在SPSS和SAS中的实现. 循证医学, 2013,13(2):120-123.
Zhou Q, Zhang J X. Data analysis for repeated measurements with missing values in SPSS and SAS.[J] Evidence-Based Med, 2013,13(2):120-123 (in Chinese with English abstract).
[9] 冯跃华, 韩钢钢, 赵田径, 董爱玲, 邹应斌, 敖和军. 单因素随机区组试验中重复测量数据分析及其SAS实现方法. 安徽农业科学, 2007,35:11730-11732.
Feng Y H, Han G G, Zhao T J, Dong A L, Zou Y B, Ao H J. Analysis of repeated measure data and its implementations of SAS in single-factor randomized complete block design. J Anhui Agric Sci, 2007,35:11730-11732 (in Chinese with English abstract).
[10] 张军锋, 董海原. 医学论文审稿中常见的统计学错误: 重复测量方法的误用分析. 中国药物与临床, 2017,17:1875-1876.
Zhang J F, Dong H Y. Common statistical errors in medical papers: Mis-use of repeated measure method. Chin Remedies Clinics, 2017,17:1875-1876 (in Chinese).
[11] 施红英, 沈毅. 混合模型在临床试验重复测量资料中的应用. 中国卫生统计, 2007,24(2):140-142.
Shi H Y, Shen Y. Mixed model applications to the analysis of repeated measures data in clinical trials. Chin Health Statistics, 2007,24(2):140-142 (in Chinese).
[12] 余松林, 向慧云. 重复测量资料分析方法与SAS程序. 北京: 科学出版社, 2004. pp 1-252.
Yu S L, Xiang H Y. The Analyse Method and SAS Procudures for Repeated Measure. Beijing: Science Press, 2004. pp 1-252(in Chinese).
[13] SAS Institute. SAS Course Notes on Mixed Model Analysis Using SAS System. Cary, NC, USA: SAS Institute Inc., 2002. pp 120-155.
[14] HJ 636-2012水质总氮的测定——碱性过硫酸钾消解紫外分光光度法. 国家环境保护标准, 2012.
HJ 636-2012 Water quality—Determination of total nitrogen-Alkaline potassium persulfate digestion UV spectrophotometric method. National Standard for Environmental Protection, 2012 (in Chinese).
[15] Littell R C, Stroup W W, Freund R J. SAS System for Linear Models, 4th edn. Cary, North Carolina, USA: SAS Institute Inc., 2002. pp 265-304.
[16] Kenward M G, Roger J H. Small sample inference for fixed effects from restricted maximum likelihood. Biometrics, 1997,53:983-997.
[17] 胡良平, 郭辰仪. 用SAS软件实现具有一个重复测量的单因素设计定量资料的统计分析. 药学服务与研究, 2011,11:409-411.
Hu L P, Guo C Y. Statistical analysis and SAS solutions for quantitative data in single-factor repeated-measurement design. Pharm Care Res, 2011,11:409-411 (in Chinese).
[18] Kiernan K, Tobias R, Gibbs P, Tao J. Contrast and Estimate Statements Made Easy: The LSMESTIMATE Statement. SAS Global Forum. Cary, North Caroina, USA: SAS Institute Inc., 2011. pp 1-19.
[19] Wolfinger R D. Heterogeneous variance covariance structures for repeated measures. J Agric Biol Environ Statistics, 1996,1:205-230.
[20] Guerin L, Stroup W W. A simulation study to evaluate PROC MIXED analysis of repeated measures data. In: Stroup W W eds. Proceedings of the 12th Annual Conference on Applied Statistics in Agriculture. Manhattan, KS: Kansas State University, 2000. pp 170-203.
[21] Huynh H, Feldt L S. Conditions under which mean square ratios in repeated measurements designs have exact F-distributions. J Am Statistical Assoc, 1970,65:1582-1589.
[22] Huynh H, Feldt L S. Estimation of the Box correction for degrees of freedom from sample data in the randomized block and split plot designs. J Education Statistics, 1976,1:15-51.
[23] 樊俊, 谭军, 邓建强, 彭五星, 赵秀云, 郑海洲, 朱宗第. 不同地膜的降解性能及对烟株生长和土壤环境的影响. 中国烟草科学, 2019,40(4):22-29.
Fan J, Tan J, Deng J Q, Peng W X, Zhao X Y, Zheng H Z, Zhu Z D. Degradation property of degradable films and their effects on flue-cured tobacco development and soil ecological environment. Chin Tob Sci, 2019,40(4):22-29 (in Chinese with English abstract).
[24] 侯红乾, 林洪鑫, 刘秀梅, 冀建华, 刘益仁, 蓝贤瑾, 吕真真, 周卫军. 长期施肥处理对双季晚稻叶绿素荧光特征及籽粒产量的影响. 作物学报, 2020,46:280-289.
Hou H Q, Lin H X, Liu X M, Ji J H, Liu Y R, Lan X J, Lyu Z Z, Zhou W J. Influence of long-term fertilizer application on chlorophyll fluorescence characteristics and grain yield of double cropping late rice. Acta Agron Sin, 2020,46:280-289 (in Chinese with English abstract).
[25] 贾国涛, 杨永锋, 杨欣玲, 王宝林, 刘超, 王根发, 申洪涛, 张书伟, 刘向真, 赵森森. 腐熟秸秆对植烟土壤理化性质和酶活性的影响. 中国农业科技导报, 2018,20(9):138-145.
Jia G T, Yang Y F, Yang X L, Wang B L, Liu C, Wang G F, Shen H T, Zhang S W, Liu X Z, Zhao S S. Influence of rotten straw on physical and chemical properties and enzyme activity of soil. J Agric Sci Technol, 2018,20(9):138-145 (in Chinese with English abstract).
[26] 陈峰, 姚晨, 孙高, 任仕泉, 何清波, 苏炳华, 陆守曾. 新药临床试验中重复测量资料的混合效应模型. 中国卫生统计, 2000,17(6):373-376.
Chen F, Yao C, Sun G, Ren S Q, He Q B, Su B H, Lu S Z. Mixed effect model of repeated measurement data in clinical trials of new drugs. Chin J Health Statistics, 2000,17(6):373-376 (in Chinese with English abstract).
[27] 王琪, 胡良平 . 如何用 SAS 软件正确分析生物医学科研资料: XIV. 用SAS软件实现具有一个重复测量的单因素设计定量资料的统计分析. 中国医药生物技术, 2011,6(4):313-320.
Wang Q, Hu L P. How to correctly analyze biomedical research data with SAS software: XIV. Statistical analysis for quantitative data with single factor design and repeated measure by using SAS software. Chin Med Biotechnol, 2011,6(4):313-320 (in Chinese with English abstract).
[28] 王超, 王汝芬, 张淑娴. 混合效应线性模型与单因素方差分析在重复测量数据中的应用比较. 数理医药学杂志, 2006,19(4):355-357.
Wang C, Wang R F, Zhang S X. The application and comparison of mixed effects linear model with single effects of variance in repeated measures data. J Mathematical Med, 2006,19(4):355-357 (in Chinese with English abstract).
[29] Littell R C, Pendergast J, Natarajan R. Modelling covariance strucutre in the analysis of repeated measures data. Statistics Med, 2000,19:1793-1819.
[1] 王娟, 张彦威, 焦铸锦, 刘盼盼, 常玮. 利用PyBSASeq算法挖掘大豆百粒重相关位点与候选基因[J]. 作物学报, 2022, 48(3): 635-643.
[2] 卞云龙,杜凯,王益军,邓德祥. 玉米茎秆糖含量的分布[J]. 作物学报, 2009, 35(12): 2252-2257.
[3] 张晓丽;郭辉;王海岗;吕建珍;袁筱萍;彭锁堂;魏兴华. 中国普通野生稻与栽培稻种SSR多样性的比较分析[J]. 作物学报, 2008, 34(04): 591-597.
[4] 王一平;魏兴华;华蕾;袁筱萍;余汉勇;徐群;汤圣祥. 不同地理来源旱稻种质资源的遗传多样性分析[J]. 作物学报, 2007, 33(12): 2034-2040.
[5] 郑天清;徐建龙;傅彬英;高用明;Satish Veruka;Renee Lafitte;翟虎渠;万建民;黎志康. 遗传搭车与方差分析在水稻定向选择群体的抗旱性位点分析中的初步应用[J]. 作物学报, 2007, 33(05): 799-804.
[6] 姜长鉴;朱庆森;邱泽森. 水平不固定的试验结果的统计分析Ⅰ.一个因素的水平不固定时的分析[J]. 作物学报, 1994, 20(02): 229-234.
[7] 李维明;吴为人;卢浩然. 检测作物数量性状基因与遗传标记连锁关系的方差分析法及其应用[J]. 作物学报, 1993, 19(02): 97-102.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!